A theoretical model traces how quantum entanglement can be represented across several geometric surfaces at once, including boundaries, end-of-the-world branes and horizons. In selected models of curved space, the construction turns these contributions into a single thread-like flow that obeys the mathematical rules expected of a valid bit-thread description.
The supplied record identifies the work as an arXiv preprint, version one, dated 26 Aug 2026, with no journal or peer-review status reported there. It uses analytic calculations rather than observations or experiments.
Turning many thread sources into one flow
The study asks whether a combined picture of “partial entanglement entropy” and bit threads can describe a gravitational subregion whose state-supporting surface has multiple components. Partial entanglement entropy, or PEE, is used here to describe how entanglement is apportioned between pieces of a boundary or auxiliary surface; bit threads are the corresponding flow lines through the geometry.
The authors construct PEE data from geodesics, match the associated fluxes, orient the thread contributions and superpose them. They then check whether the resulting current has no unwanted sources or sinks, stays below the allowed pointwise norm, and reaches the relevant bottleneck surface with the required saturation. In plain terms, the test asks whether separately sourced pieces can be added without creating an invalid flow.
The calculations cover Poincaré AdS3, planar BTZ black branes with end-of-the-world branes, and entanglement wedges associated with boundary multi-intervals. These are analytic configurations rather than a statistical sample, and the paper does not use statistical inference.
What changes when a horizon is included
In the BTZ construction, the extended decomposition includes boundary-to-boundary, boundary-to-brane, brane-to-brane, boundary-to-horizon and brane-to-horizon channels. The horizon is treated as an auxiliary thermal purifier in this bookkeeping, rather than as a physical boundary of the boundary theory.
There is an important qualification. For separated points on the horizon, the smooth horizon-to-horizon PEE kernel vanishes, so the analysis does not produce a separate smooth bulk source field for that channel. Contributions at coincident points are retained only as contact data.
The extra sectors change how the entanglement contribution is divided up, but the connected physical answer remains correctly normalized in the geometries examined. In phases where an interval ends on the brane, the extended construction reproduces the Affleck–Ludwig boundary entropy. For connected intervals away from the boundary, brane-dependent pieces cancel and the ordinary conformal-field-theory result remains.
The flow need not follow the shortest paths
One of the study’s more counterintuitive findings is that the final bit-thread streamlines are generally not geodesics, even though the elementary PEE threads used to build them are. Where the combined current is smooth and nonzero, its streamlines form a regular, non-crossing foliation of the region.
For the Poincaré AdS3/BCFT2 cases, the boundary- and brane-sourced contributions form a divergenceless current whose magnitude stays within the bound |v| ≤ 1/(4G), and the flow saturates on the brane-ending Ryu–Takayanagi geodesic. The planar-BTZ fields are derived from BTZ geodesics, with the boundary field obtained by an isometry push-forward; the resulting currents pass the stated flux and RT-entropy checks in the chambers studied.
A recurring simplification is that shared auxiliary endpoints cancel when the source flows are integrated across the extended surface. What remains is controlled by the relative boundary of the complete extended subsystem, rather than by every internal source endpoint separately.
Using geometry to choose a purifier
The same framework is then used to study geometric minimal purification within the paper’s surface/state class. The authors combine PEE threads sourced from the purified subsystem with its auxiliary purifier, producing a divergenceless max-flow that saturates on the candidate entanglement-wedge cross section, or EWCS.
In the Poincaré and planar-BTZ examples, integrating the sources reduces to an endpoint-potential expression. Its flux equals the length of the candidate cross section, and optimizing how the purifier is split selects the EWCS. This gives a geometric realization of the paper’s EP–EW correspondence within the specified class of purifications.
For symmetric disjoint intervals in planar BTZ, the connected-phase optimum is the unique common perpendicular at x = 0. When ηβ ≤ 1/2, the dominant wedge is disconnected and the leading classical EWCS vanishes.
A construction with a defined boundary
The results are limited to the explicitly analyzed geometries and phases. The paper gives explicit AdS/BCFT flow formulae for adjacent brane-ending and connected chambers, while disconnected BCFT phases have an intersection-weighted normalization but no explicit PEE-generated max-flow representative here.
The interpretation of PEE densities on branes, horizons and RT surfaces as microscopic entanglement data is conditional on an extended surface/state dictionary. The geometric purification result is also limited to that surface/state class and does not establish minimization over the full quantum-information-theoretic space of purifications.
Taken together, the calculations support geometric consistency in the selected chambers, while leaving open whether the framework can be made explicit in disconnected BCFT phases. The paper’s normalization covers those phases through intersection weighting, but its explicit PEE-generated flow representative remains an open part of the construction.
Paper data and sources
Original title: PEE threads and bit threads in gravitational subregions of AdS
Authors: Debarshi Basu, Qiang Wen
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text